The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection by ingesting untrusted data from external sources like GitHub READMEs and paper collections.
Ingestion points: Data is retrieved from GitHub repositories, paper lists, and web search results as described in the search and screen sections of SKILL.md.
Boundary markers: The skill explicitly requires using the paper-card-delivery skill as a 'Canonical Paper Card Gate' for official-source verification and fixed card formatting, which acts as a validation boundary.
Capability inventory: The skill generates research artifacts and uses external adapters (notion-doc-workflow, feishu-doc-workflow, etc.) to write content to external documentation platforms.
Sanitization: The workflow mandates that technical claims must be followed up with official paper sources (PDF/Project/Code) before writing reader-facing content, reducing the risk of malicious data influence.

survey-builder